The early part of the chart shows some alignment with estimated actual mean temperatures in the lower troposphere derived from satellite observations (UAH, RSS), but this is not due to the predictive skill of the models so much as the parameter adjustment skills of the modellers trying to get good fits to past observations. Spencer does not report on when the forecasting part of the model outputs begin on these plots, but in the comments below his post he guesses that it may be no later than 2007. Ignoring the 1998 spike widely attributed to a very strong El Nino, the sustained divergence of actual from model seems to begin in about 1995.
Note also that there are other forecasts of global mean temperatures that have a better track record so far than the climate models. Here are two:
(1) Global surface temperatures projected from 2007 model runs are shown in the green band. The blue (cyan) band is for the empirically-fitted model by Scafetta in which he merely makes a combination of past, observed cycles in temperature plus ad adjustment to reflect the overall warming of the 20th century. The thick red line changing to thick blue is derived from observations. The thick black line shows Scafetta's model projected into the future. So far, it is doing a lot better than the multi-million pound GCMs. I reckon Scafetta's model could be run on a decent programmable calculator such as this one costing about £30.
(2) An even less expensive prediction model is the basic persistence one devised by an expert in forecasting techniques after working through a checklist of what it takes to produce forecasts based on best practices (a checklist against which the climate forecasts of such as the IPCC fail dramatically). This expert was so dismayed by the poor basis for forecasts made by the IPCC and taken up by such as Al Gore, that he announced a public bet over whether Gore's claim of a 3C rise by the end of the 21st century (and this is at the moderate end of IPCC predictions) could beat a simple persistence forecast. In the 63 months since the bet began, the persistence forecast has done better in 55 of them. Details are here: The Global Warming Challenge.
These plots can help your pupils regard the IPCC, and assorted CO2 alarmists in general with the contempt they deserve. They may have to regurgitate their 'science' to pass exams, but they can treat it like theology rather than science - the exam answers are then more about the faith of the alarmed ones than about science or the world outside of their expensive but woefully inadequate General Circulation Models (GCMs).
Note added 20 April 2013 Taking the model outputs seriously, because to do so suits them very much, alarmed ones have made many blunders. Pierre Gosselin has assembled a collection relating to their confident assurances re warmer winters for Europe. Examples can be seen here: http://climatelessons.blogspot.co.uk/2013/04/climate-teachers-have-you-seen-any-of.html